patients.R

#https://archive.ics.uci.edu/ml/datasets/Post-Operative+Patient
library(neuralnet)
library(dplyr)

read.table('post-operative.data',sep=',') -> po
randomized_data <- sample_frac(po,1)

training <- randomized_data[1:79,]
testd <- randomized_data[80:90,]

training <- mutate(training,
                   A = V9 == "A",
                   S = V9 == "S",
                   I = V9 == "I",
                   highCore = V1 == "high",
                   midCore = V1 == "mid",
                   lowCore = V1 == "low",
                   highSurf = V2 == "high",
                   midSurf = V2 == "mid",
                   lowSurf = V2 == "low",
                   excellent = V3 == "excellent",
                   good = V3 == "good",
                   fair = V3 == "poor",
                   poor = V3 == "poor",
                   highBp = V4 == "high",
                   midBp = V4 == "mid",
                   lowBp = V4 == "low",
                   stableSurf = V5 == "stable",
                   modStableSurf = V5 == "mod-stable",
                   unstableSurf = V5 == "unstable",
                   stableCore = V6 == "stable",
                   modStableCore = V6 == "mod-stable",
                   unstableCore = V6 == "unstable",
                   stableBp = V7 == "stable",
                   modStableBp = V7 == "mod-stable",
                   unstableBp = V7 == "unstable"
)
nn <- neuralnet(
  A + S + I ~ highCore + midCore + lowCore + highSurf + midSurf + lowSurf + excellent + good + fair + poor + highBp + midBp + lowBp + stableSurf + modStableSurf + unstableSurf + stableCore + modStableCore + unstableCore + stableBp + modStableBp + unstableBp,
  data = training,
  hidden = c(10,10),
  act.fct = "logistic",
  linear.output = FALSE
)
plot(nn)
Predict=compute(nn,testd)
res = Predict$net.result
res